[Bioclusters] Blast Problems

Joe Landman bioclusters@bioinformatics.org
Mon, 23 Feb 2004 07:49:39 -0500


On Mon, 2004-02-23 at 07:24, david speed (RI) wrote:
> Hi all,
> 
> We have installed Sun Grid Engine onto our 15-node Linux (Red Hat 7.1)
> cluster (30 Intel CPUs). There is an NFS export mounted from the head
> node to each slave node solely to contain the SGE tools and
> directories.  We have installed the ncbi blast tools (blastall version
> 2.2.6) and the databases to be blasted against locally on each node.
> 
> This set up works fine for most of the jobs we run however sometimes a
> blast job will continue to run even though the output file contains
> output for all the sequences in the input file the job just runs using
> 99.9 %cpu until the queue is cleaned.  After the queue is cleaned
> sometimes the queue will become unusable.

This could be the run blocking for some reason.  Does this occur
randomly in the cluster or on only one or two nodes?

Does it happen with the newer tools?  You can grab RPMs for the newer
tools from http://downloads.scalableinformatics.com/downloads/ncbi/ 


> For a long time we thought this was a problem with our SGE set-up now
> we are thinking it may be a problem with ncbi blast has anyone else
> had similar problems with blast failing to exit?

If the process is hung, try running strace against it.  

	strace -p PID_of_hung_blast

to see where it is blocking.

> 
> Would there be any benefit in switching to Wu-Blast?

Possibly, though I would imagine that this would be the case for the
differences in implementation of the algorithms rather than as a
"bugfix" case vs NCBI.

The 2.2.7 tools are out (and in theory the 2.2.8).  I would suggest that
you try them, and see if they fix the problem.

Joe

> 
> Thanks for your time
> 
> David Speed
> Programmer
> Roslin Institute
> Bioinformatics Group
> Roslin, 
> Midlothian, 
> EH25 9PS, 
> UK
> Telephone: +44 (0)131 527 4200 (switchboard) 
> Fax: +44 (0)131 440 0434
> 
> The information contained in this e-mail (including any attachments) is confidential and is intended for the use of the addressee only. The opinions expressed within this e-mail (including any attachments) are the opinions of the sender and do not necessarily constitute those of Roslin Institute (Edinburgh) ("the Institute") unless specifically stated by a sender who is duly authorised to do so on behalf of the Institute.
> 
> 
> _______________________________________________
> Bioclusters maillist  -  Bioclusters@bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/bioclusters